The spectrum of bond-forming reactions catalyzed by naturally occurring enzymes, e.g., naturally occurring sortases, ligases, polymerases, and kinases, is limited and typically restricted to specific substrates. Such enzymes can be used to form bonds between molecules, e.g., proteins, nucleic acids, carbohydrates, or small molecules, under physiological conditions, thus allowing in vivo and in vitro modification of molecules in or on living cells and other biological structures while maintaining their structural integrity. For example, sortases catalyze a transpeptidation reaction that results in the conjugation of a peptide comprising a C-terminal sortase recognition motif with a peptide comprising an N-terminal sortase recognition motif. Naturally occurring sortases are typically selective for specific C-terminal and N-terminal recognition motifs, e.g., LPXTG (SEQ ID NO: 2; where X represents any amino acid) and GGG, respectively. The T and the G in the substrate can be connected using a peptide bond or an ester linkage. The spectrum of peptides and proteins that can be conjugated via sortases is, therefore, limited. While target proteins not comprising a sortase recognition sequence may be engineered to add such a sequence, such engineering is often cumbersome or impractical, e.g., in situations where the addition of an exogenous sortase recognition motif would disturb the structure and/or the function of the native protein. Another obstacle to a broader application of bond-forming enzymes to biological systems is that naturally occurring bond-forming enzymes typically exhibit low reaction efficiencies. The generation of bond-forming enzymes that efficiently catalyze bond-forming reactions and/or utilize a different, non-natural target substrate, e.g., a desired sortase recognition sequence, would allow for a broader use of sortases to modify proteins in research, therapeutic, and diagnostic application.